Each test in the Divergent Production battery requires the examinee to produce a response. Since these responses must be evaluated, the factor of rater judgment influences the reliability of scores. The problem of scoring reliability is one which pervades the literature on creativity research, where either low estimates or no estimates have been reported when tests from the battery are used. The purpose of this study was to develop a training program for raters of some Divergent Production fluency factor tests and to evaluate this program. An experimental design was generated for the evaluation and the scoring reliability was estimated through analysis of variance procedures. General principles for training raters and for analyzing the results of the design will be discussed. (Author)